Alternative Etiologies of Liver Disease in Children With Suspected NAFLD

This was a multicenter, retrospective cohort study of patients aged ≤18 years with a BMI at or above the 85th percentile for age referred for suspected NAFLD to the steatohepatitis clinic at Cincinnati Children’s Hospital Medical Center from 2009 to 2017 or the hepatology clinic at Yale New Haven Children’s Hospital from 2012 to 2017. NAFLD was suspected on the basis of either elevation of serum aminotransferase levels or imaging consistent with hepatic steatosis (liver ultrasound, MRI and/or computed tomography scan). Patients with known preexisting liver diseases before the initial clinic visit, those who reported any alcohol consumption, and those with baseline serum ALT levels >500 U/L were excluded. Institutional Review Board approval and a waiver of informed consent were obtained at both institutions before the initiation of data collection.

As per the North American pediatric NAFLD guidelines,¹  standardized testing was performed to evaluate for the following conditions: AIH, Wilson disease, viral hepatitis B and C, thyroid dysfunction, celiac disease, A1AT deficiency, and hemochromatosis (as detailed under Screening for Other Chronic Liver Diseases below). Patients received standard-of-care dietary, weight management, and physical activity recommendations, including advice to stop sugar-sweetened beverage consumption, decrease meal portion sizes, increase fruit and vegetable consumption, optimize their physical activity levels, and improve their sleep habits.^1,20  Some patients subsequently underwent a liver biopsy per clinical indications. Typical indications for a liver biopsy included concerns for one of the following: (1) severe NAFLD (persistently elevated ALT level [>50 U/L] for 3–6 months and lack of improvement after appropriate reported lifestyle changes [weight loss or improvements in BMI]; cardiometabolic risk factors associated with more severe NAFLD, such as T2DM or OSA; splenomegaly and/or increased liver stiffness on magnetic resonance elastography²¹ ) or (2) another underlying liver disease (eg, AIH due to positive autoantibodies with elevated serum immunoglobulin G levels).

Electronic medical records were reviewed to collect clinical characteristics (age, sex, ethnicity, anthropometrics) from the first clinic visit and laboratory data obtained within 6 months of first clinic visit (whichever was closest), including serum levels of ALT, aspartate aminotransferase (AST), γ-glutamyltransferase (GGT), alkaline phosphatase (ALP), fasting glucose and insulin, hemoglobin A1c (HbA1c), homeostasis model assessment of insulin resistance (HOMA-IR), and lipid profile.

A normal ALT level was defined as ≤22 IU/L in girls and ≤26 IU/L in boys.²²  Using Centers for Disease Control and Prevention growth charts, we classified the severity of obesity as follows: overweight: BMI 85th to <95th percentile for age and sex; class I obesity: BMI 95th to <120% of the 95th percentile; class II obesity: BMI 120% to <140% of the 95th percentile; class III obesity: BMI ≥140% of the 95th percentile.²³  We determined the presence of T2DM on the basis of either a HbA1c >6.4%, an oral glucose tolerance test with a plasma glucose level >200 mg/dL at 2 hours, or an established T2DM diagnosis by an endocrinologist.²⁴ 

Serological tests for chronic liver diseases (antinuclear antibody [ANA], anti–smooth muscle antibody [ASMA], anti–liver-kidney microsomal antibody [anti-LKM], immunoglobulin G, A1AT level, protein plasma isoelectric [Pi] typing, ceruloplasmin, immunoglobulin A [IgA], tissue transglutaminase [tTG]–IgA, ferritin, and free thyroxine (T4) or thyroid-stimulating hormone [TSH]) were also collected from the electronic medical records.

The diagnosis of AIH was based on a combination of clinical, biochemical, immunologic, and histologic features per previously published guidelines.²⁵  All 3 antibodies (ANA, ASMA, and anti-LKM) were measured by indirect immunofluorescence by using human epithelial type 2 cells for ANA and mouse liver, kidney, or stomach tissue for ASMA and anti-LKM. ANA was defined by the commercial technique as ≥1:80, whereas an undetectable ANA was reported as <1:80. Any ASMA and anti-LKM positivity was reported according to the commercial techniques, which differed at each institution: ASMA level ≥1+ or elevated F-actin ASMA level >20 and anti-LKM level ≥1+ or any detectable antibody titers.

A serum ceruloplasmin level <20 mg/dL was considered low.²⁶  Subjects with low ceruloplasmin levels either underwent a 24-hour urine collection for copper quantification or proceeded directly to a liver biopsy with determination of hepatic tissue copper concentration. Urinary copper excretion >40 μg per 24 hours was considered excessive and was followed by a confirmatory liver biopsy with copper quantification.²⁶ 

Serum TSH was used to test for thyroid dysfunction.^27,28  A free T4 test was obtained either routinely with the TSH or as a reflex, with the T4 measured only when the TSH was abnormal. Abnormal serum TSH levels and free T4 levels were defined by using institutional and age-specific reference ranges.

Patients with elevated tTG-IgA levels were investigated further with upper endoscopy.²⁹  However, those with normal tTG-IgA levels in the setting of low total IgA did not always undergo an endoscopy because the potential risks of sedation were deemed higher than the likelihood of histologic confirmation in children with severe obesity and potential OSA.

A1AT protein phenotype testing (π typing) or A1AT levels were measured to test for A1AT deficiency. Positive π tying was defined as SZ, MS, MZ, or ZZ phenotypes.^30,31 

The screen result for hemochromatosis was considered positive if the serum transferrin saturation was ≥45% along with serum ferritin levels >300 μg/L or if the serum ferritin levels were >500 μg/L.³²  These findings were then followed by MRI of the liver with iron quantification for further confirmation of iron overload and/or a liver biopsy.

For those who underwent a clinically indicated liver biopsy, hepatic histology was scored by experienced hepatopathologists using the classification developed by the nonalcoholic steatohepatitis (NASH) Clinical Research Network: steatosis score 0 to 3, lobular inflammation score 0 to 3, hepatocyte ballooning score 0 to 2, and fibrosis stage 0 to 4. The NAFLD activity score was calculated as the sum of the scores for steatosis, lobular inflammation, and ballooning (range 0–8).³³  Definite NASH was diagnosed by a pattern of liver injury, including all 3 components of hepatic steatosis, lobular inflammation, and ballooning degeneration.³⁴ 

Descriptive statistics (medians and interquartile ranges [IQRs] for continuous variables and frequencies and percentages for categorical variables) were used to present the demographics and clinical characteristics of the cohort. χ² testing was used to compare normal categorical variables, and the Mann–Whitney U test was used to compare continuous variables between patients with biopsy-confirmed NAFLD and those with presumed NAFLD once those with other liver diseases were excluded. Univariate and multivariable regression analyses were used to further identify independent variables associated with biopsy-confirmed versus presumed NAFLD. All variables found to be significantly associated with the groups in the univariate analyses were included in a multivariable model. Analyses were performed by using Stata MP v.14.2 (Stata Corp, College Station, TX). Significance was set as P ≤ 0.05.

A total of 900 children with overweight or obesity were referred for suspected NAFLD during the study periods. Their median age was 13 years (IQR: 10–16); 63% of the patients were boys, and 26% were Hispanic. The majority had severe obesity (n = 666; 76%). The median ALT level of the cohort at the first visit was 64 U/L (IQR: 42–95). The ALT level was normal in 5% of patients, 1 to 2 times the ULN in 29%, and >2 times the ULN in 66%. The remaining demographic baseline clinical and laboratory characteristics of the cohort are summarized in Table 1.

The results of the exclusionary testing performed are revealed in Fig 1. Of the 900 patients with suspected NAFLD, a total of 19 (2%) were diagnosed with other chronic diseases (as revealed in Fig 1, these were the following: celiac disease, n = 3; thyroid dysfunction requiring treatment, n = 11; A1AT, n = 3; Hodgkin’s lymphoma, n = 1, and hemophagocytic lymphohistiocytosis, n = 1). There were no patients with ≥2 additional diagnoses picked up with the exclusionary testing. Positive autoantibodies for AIH were observed in 13% of patients (95 of the 734 tested); however, none met composite criteria for AIH. No patient was diagnosed with viral hepatitis (B [0 of the 555 tested] or C [0 of the 572 tested]) or Wilson disease (0 of the 736 tested).

As revealed in Table 1 and Fig 1, 702 patients were screened for celiac disease. Five percent were found to be IgA-deficient, and 1% had evidence of elevated tTG-IgA levels. Ultimately, 3 patients (0.4%) proceeded to have confirmed celiac disease. From a thyroid perspective, 630 patients were screened for thyroid dysfunction and 14% (n = 88) had abnormal TSH levels. Ultimately, thyroid dysfunction that required treatment by an endocrinologist was found in 11 of those patients. Seven hundred twenty-one patients were tested for A1AT deficiency by using phenotyping and/or serum A1AT levels. Three of them (0.4%) were ultimately diagnosed with A1AT deficiency. Lastly, initial testing for hemochromatosis was performed by using serum ferritin levels (n = 622) and/or transferrin saturation (n = 563), and, ultimately, no patient was diagnosed with this condition.

Of the 881 patients who did not have other diseases (ie, excluding the 19 patients who were diagnosed with other or concurrent diseases) a total of 347 children (39%) had biopsy-confirmed NAFLD. Of those, 46% had NASH, as revealed in Table 2.

Compared with those with presumed NAFLD, the biopsy-confirmed NAFLD cohort had a higher proportion of patients of Hispanic ethnicity or with T2DM. In addition, serum levels of ALT, AST, GGT, cholesterol, low-density lipoprotein cholesterol (LDL-C), and triglycerides, as well as markers of glucose intolerance, were significantly higher in those who had undergone a liver biopsy (Table 2).

In this large, multicenter pediatric cohort, we found that the vast majority of children with overweight and obesity BMI referred for suspected NAFLD tested negative for other causes of liver disease or liver enzyme elevation. In contrast to a previous report,¹⁹  no patient in our cohort was diagnosed with AIH. A small number of patients with confirmed NAFLD and/or NASH were also found to have concurrent celiac disease, A1AT deficiency, or thyroid dysfunction; however, they continued to be managed and treated for NAFLD in addition to their secondary diagnosis. This finding highlighted that NAFLD can coexist with other conditions also known to cause elevations in liver enzymes.

When caring for patients with suspected NAFLD, it remains critical to evaluate for concurrent other chronic diseases that may mimic or coexist with NAFLD because some of these conditions require specific pharmacotherapy and/or may progress rapidly without appropriate treatment. One such condition is AIH, which has a reported prevalence of 3 per 100 000 children in the United States.²⁵  Schwimmer et al¹⁹  found that 11 of the 255 (4%) children with overweight and obesity BMI referred to their center for suspected NAFLD had evidence of AIH, rendering AIH the second most common diagnosis, after NAFLD, in that population. In contrast, none of the 900 children referred for suspected NAFLD to our centers met diagnostic criteria for AIH, which aligns with the low prevalence of AIH in the general population. These discrepant findings may potentially reflect race and ethnicity differences between the 2 study cohorts. This remains to be investigated further in larger population and cost-effectiveness studies. In the meantime, it is important to continue testing for AIH in patients with suspected NAFLD because early identification and treatment of AIH is important for future outcomes.³⁵ 

Another treatable liver disease that is important to exclude in patients with suspected NAFLD is Wilson disease. Wilson disease, a condition of excessive copper accumulation, has a prevalence of 1 in 30 000 in the general population and can present with moderate to severe hepatic steatosis in 50% of children, which is the earliest change seen in the liver pathology.^36,37  The clinical manifestations of Wilson disease are diverse and can range from complete absence of symptoms to subtle neurologic disease to overt liver disease and even failure.³⁸  Similar to NAFLD,³⁹  modest elevations in ALT (100–500 U/L) and GGT levels may be the sole feature at presentation in Wilson disease.³⁸  In addition, the histologic appearances of NAFLD and Wilson disease can be indistinguishable with routine histologic stains. Early treatment is critical to improve long-term outcomes and prevent progression to end-stage liver disease in patients with Wilson disease.³⁸  Considering the similarities in the clinical presentation, the availability of treatment, and the risk of rapid progression without treatment, Wilson disease is an important diagnosis to exclude in patients with suspected NAFLD. The inability to detect any patients with Wilson in our cohort is likely reflective of the discrepancy between our sample size and the overall prevalence of Wilson and by no means suggests that testing practices for this condition in patients with suspected NAFLD should change.

In the United States, the prevalence of celiac disease ranges from 1 in 80 to 1 in 300 children.⁴⁰  In addition to the known risk of developing celiac hepatopathy, patients with celiac disease may also be at increased risk of NAFLD, which may in part reflect excess weight gain that can occur when consuming a highly processed gluten-free diet.⁴¹  In contrast, the data on whether patients with other chronic liver diseases are at an increased risk of celiac disease are conflicting.^41–45  Most studies in the field have been done on adults and not on children or adolescents with obesity. In one pediatric study, none of the 120 children with obesity undergoing evaluation for NAFLD had positive testing results for celiac disease.²⁹  Consensus is lacking on whether testing for celiac disease should be undertaken when assessing asymptomatic children for NAFLD. In our study, 3 of 900 patients had confirmed celiac disease, which parallels the prevalence seen in the general population. It should be noted, however, that IgA deficiency, which is a risk factor for celiac disease,²⁹  was found in 5% of our patients. Only 40% of this subgroup was investigated further with endoscopy because they were asymptomatic, and sedation for endoscopy was considered higher risk in the context of severe obesity and related comorbidities. Thus, there is the potential for underestimation of the true prevalence of celiac disease in our cohort. Importantly, when celiac disease is concurrently diagnosed in the context of NAFLD, ongoing monitoring of the liver disease and dietary and lifestyle modification is required because both conditions require lifelong management.

Thyroid abnormalities were detected in 1.2% (11/900) of our cohort. Thyroid dysfunction is common in obesity: 7% to 23% of pediatric patients with obesity have elevated TSH levels with normal free T4 and triiodothyronine (subclinical hypothyroidism).^46,47  According to a recent meta-analysis, true hypothyroidism occurs with a prevalence of 14% in subjects with obesity.⁴⁸  Hypothyroidism is important in the context of NAFLD because it attenuates fatty acid oxidation and triglyceride export from the hepatocytes, favoring the development of hepatic steatosis.⁴⁹  Supporting this relationship, recent data from an Italian cohort of 403 children with overweight- or obesity-range BMI revealed that elevated TSH levels (>4 mIU/L) were associated with increased odds of hepatic steatosis.⁵⁰  A subsequent study of 332 children with overweight- or obesity-range BMI from Germany revealed that patients with NAFLD had significantly higher TSH levels than their counterparts with obesity without NAFLD and that higher TSH quartiles were associated with a greater likelihood of ultrasound-determined steatosis.⁵¹  To date, it is not clear whether thyroid hormone replacement therapy for patients with hypothyroidism can ameliorate NAFLD clinically or histologically.⁴⁹  In no study has the effect of treatment of true hypothyroidism on NAFLD been examined. The results of small trials aimed at treating subclinical thyroid dysfunction to improve NAFLD in adults have generated conflicting results.^52–54  Understanding thyroid dysfunction in NAFLD is important because it may identify subphenotypes of patients with potentially different outcomes and treatment options.

The prevalence of A1AT deficiency in the United States is between 1 in 2857 and 1 in 5097.⁵⁵  Although it is currently not clear what the prevalence of A1AT deficiency is in children or adolescents with obesity undergoing investigations for presumed NAFLD, a much higher number than what would be expected on the basis of general population data, namely 3 of 900 (or 1 in 300), had A1AT deficiency in our cohort. The overrepresentation of A1AT phenotypes in NAFLD has been previously reported in adult cohorts. Specifically, Pi*Z and Pi*S heterozygosity have been revealed to increase the risk of advanced liver disease in this context.^56,57  Strnad et al⁵⁶  found that the risk of cirrhosis among adults with NAFLD was increased by more than sevenfold in heterozygous carriers of the A1AT PI*Z variant. Similarly, Z phenotypes have been found to be overrepresented in patients with NAFLD undergoing a liver transplant in Ireland.⁵⁸  Although more pediatric data on the topic are needed, the existing literature supports the role of π typing for patients with suspected and/or confirmed NAFLD, not only as a diagnostic, but also as a risk-stratification tool.

In our study, approximately one-half of the children underwent a liver biopsy to confirm the presence of NAFLD. As expected, the biopsy-confirmed NAFLD cohort had higher ALT values than those with presumed NAFLD, as well as higher AST, GGT, serum insulin, HOMA-IR, cholesterol, LDL-C, and triglyceride levels. This is likely reflective of our clinical practice, in which we obtain liver biopsies in patients with persistently elevated liver enzymes or those with concerning cardiometabolic comorbidities, such as the presence of T2DM, which has been associated with higher risk of more severe NAFLD.⁵⁹  Of those who had a liver biopsy, nearly one-half had NASH. Although the pediatric NAFLD guidelines suggest referring patients to pediatric gastroenterologists when the ALT is elevated >2 times the ULN,¹  our data suggest that one-third of our patients were referred for ALT elevations that did not meet >2 times the ULN threshold.¹  It has been previously revealed that, even among patients with NAFLD, normal ALT levels or mild ALT elevations (<2 times the ULN) do not exclude the presence of NASH or even fibrosis.⁶⁰ 

Strengths of this study include the large sample size, the multicenter design, and the inclusion of an ethnically and racially diverse patient population. Limitations include its retrospective design, which was associated with missing data. Indeed, not all 900 patients had been tested for all the alternative or secondary¹  conditions of interest, which means that alternative diagnoses may have been underestimated. It is unclear why these tests were not done. It is possible that they were not repeated at our institutions because they had been done (and results found to be negative, given the lack of clinical follow-up) at outside laboratories and hospitals; however, we did not have access to any other results to confirm this hypothesis. For certain conditions, such as celiac disease, the subsequent testing was not always definitive, as was the case with patients with IgA deficiency who did not undergo a subsequent endoscopy. Finally, liver biopsies were not obtained in the entire cohort and, as such, histologic exclusion of other liver diseases was not universally available; however, most of these other conditions can be diagnosed on the basis of serological testing without a clinical indication for liver biopsy. In addition, the inclusion of patients with biopsy-confirmed, as well as presumed, NAFLD more accurately reflects clinical practice because it included both patients who were concerning enough to require a liver biopsy and patients with a milder presentation who did not need to proceed with a biopsy. Lastly, this study did not routinely include investigations for rare conditions, such as lysosomal acid lipase deficiency or other viral causes of liver disease such as hepatitis A virus, Epstein-Barr virus, cytomegalovirus, etc. These acute viral infections were less likely in these patients, given the prolonged persistent elevation in the serum aminotransferase levels and the exclusion of patients with ALT >500 U/L.

In this multicenter, large pediatric cohort, we found that the vast majority of children with overweight- or obesity-range BMI and suspected NAFLD tested negative for other causes of chronic liver disease. In contrast to previous reports, no patient was diagnosed with AIH. Nonetheless, NAFLD remains a diagnosis of exclusion, and key conditions that require specific treatments must be ruled out in the workup of patients with suspected NAFLD. In the future, the cost-effectiveness of this approach will need to be investigated.

Pi

protein plasma isoelectric

A1AT

α-1 antitrypsin

AIH

autoimmune hepatitis

ALP

alkaline phosphatase

ALT

alanine aminotransferase

ANA

antinuclear antibody

anti-LKM

anti–liver-kidney microsomal antibody

ASMA

anti–smooth muscle antibody

AST

aspartate aminotransferase

GGT

γ-glutamyltransferase

HbA1c

hemoglobin A1c

HOMA-IR

homeostasis model assessment of insulin resistance

IgA

immunoglobulin A

IQR

interquartile range

LDL-C

low-density lipoprotein cholesterol

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

OSA

obstructive sleep apnea

T2DM

type 2 diabetes mellitus

T4

thyroxine

TSH

thyroid-stimulating hormone

tTG

tissue transglutaminase

ULN

upper limit of normal